How to configure http-autoscaler in AppHost code?

[bitbound]

I'm deploying to Azure Container Apps via azd up.

I was able to configure http-autoscaler with azd infra synth and modifying the templates, but it would be nice if there was a way to do it in AppHost code without creating the templates on disk.

Is there way to configure minReplicas, maxReplicas, and http-autoscaler concurrent requests in AppHost code? Or does it require using the templates?

[davidfowl]

It’s not supported yet. It requires dropping to bicep using azd infra synth and modifying scaling rules

[NPadrutt]

I was looking for that aswell. Concretly the minReplicas since I want to avoid unnecessary cost as I'm not yet out of an early pilot phase yet. Is it planned to support that in the future?

[BartNetJS]

@davidfowl is it on the radar?
I also like to scale to 0 to avoid costs on POC's, demo's, etc

[sveinhelge]

You can set min and max like this.

project.PublishAsAzureContainerApp((AzureResourceInfrastructure construct, ContainerApp app) =>
{
    containerApp.Template.Scale.MinReplicas = BicepHelper.ConvertToInt(ApiConfigParameters.MinReplica);
    containerApp.Template.Scale.MaxReplicas = BicepHelper.ConvertToInt(ApiConfigParameters.MaxReplica);
})

internal class BicepHelper
{
      public static FunctionCallExpression ConvertToInt(string name)
      {
          return new FunctionCallExpression(new IdentifierExpression("int"), new IdentifierExpression(GetBicepIdentifier(name)));
      }

      private static string GetBicepIdentifier(string name)
      {
          return $"{name.Replace('-', '_')}_value";
      }
      
      public static ProvisioningParameter GetBicepParameter(Aspire.Hosting.Azure.AzureResourceInfrastructure construct, string paramName)
        {
            return construct.GetProvisionableResources().OfType<ProvisioningParameter>()
                .Single(p => p.BicepIdentifier == GetBicepIdentifier(paramName));
        }
}

public class ApiConfigParameters
    {
        private const string Prefix = "api-";
        public const string MinReplica = $"{Prefix}container-min-replica";
        public const string MaxReplica = $"{Prefix}container-max-replica";
        public const string Cpu = $"{Prefix}container-cpu";
        public const string Ram = $"{Prefix}container-ram";
    }

From the config.json

{
  "infra": {
    "parameters": {
       "api_container_min_replica": "0",
       "api_container_max_replica": "3",
    }
  }
}

CPU and RAM can be done like this. Same concept as above

containerApp.Value!.Resources.Cpu = BicepHelper.GetBicepParameter(construct, ApiConfigParameters.Cpu);
containerApp.Value!.Resources.Memory = BicepHelper.GetBicepParameter(construct, ApiConfigParameters.Ram);

[davidfowl]

You can model parameters much more simply:

var builder = DistributedApplication.CreateBuilder(args);

var replicas = builder.AddParameter("minReplicas");

// Testing projects
builder.AddProject<Projects.AzureContainerApps_ApiService>("api")
       .WithExternalHttpEndpoints()
       .PublishAsAzureContainerApp((module, app) =>
       {
           // Scale to 0
           app.Template.Scale.MinReplicas = replicas.AsProvisioningParameter(module);
       });

builder.Build().Run();

[sveinhelge]

Very nice, then I can get rid of more plumbing code.
Will app.Template.Scale in the future get CooldownPeriod and PollingInterval props like we see in Bicep?

[davidfowl]

That would be a question for azure container apps. Aspire is just exposing whatever settings are supported by the service.

[VistianOpenSource]

If you install v 1.1 of Azure.Provisioning.Appcontainers you'll find properties in ContainerAppScale for CooldownPeriod and PollingInterval have been added.